In this paper, we propose an anti-interception scheme to enhance the defence performance of frequency hopping (FH) tactical systems against the in-band full-duplex (IBFD) interception from the enemy. Our scheme utilizes an unmanned aerial vehicle (UAV)-based Intelligent Reflecting Surface (IRS) to share the interception burden with a FH system, thus mitigating jamming effects. To efficiently address IBFD interception, we jointly optimize the control of base station transmit power, FH hopping decision, and IRS phase shift adjustment. This joint optimization is mathematically formulated as a Mixed Integer Programming (MIP) non-convex optimization problem. To address the intractability of traditional optimization methods in solving this problem in modern tactical scenarios, we design a solution based on deep reinforcement learning (DRL). Extensive numerical results show that our proposed scheme significantly enhances the FH anti-interception capability and improves the QoS. Furthermore, the performance of our DRL solution is close to optimal and it is feasible to be deployed in modern practical scenarios.

The frequency hopping (FH) technique has always been crucial for anti-jamming tactical applications thanks to its advantages in avoiding the jammer's interception. However, modern tactical scenarios require FH systems to not only undertake defence missions but also meet increasingly high Quality of service (QoS) requirements. Unlike the prior works that mainly optimize FH systems by balancing anti-jamming capability and QoS performance, we propose a collaboration of FH and the intelligent reflecting surface (IRS) in an advanced anti-jamming scheme. Such approach shares the burden with the IRS and improves QoS. We formulate a joint IRS-aided FH anti-jamming problem as a Mixed Integer Programming (MIP) non-convex optimization. To address the intractability of traditional optimization methods in solving this problem in modern tactical scenarios, we design a solution based on deep reinforcement learning (DRL). The numerical results show that the performance of our solution is close to optimal, and it is scalable to be applicable in practical situations.

In this paper, we design a jamming mitigation plan to protect a mixed radio frequency/free-space optical (RF/FSO) relay network in the context that both RF and FSO systems are simultaneously attacked by enemy jammers. Our design aims to jointly optimize the power allocation (PA) and Field-of-View (FoV) tuning strategy to maximize the RF uplink sum rate subject to practical constraints on the jamming mitigation in both FSO and RF systems. In order to address the underlying non-convex optimization problem, we first derive the closed-form expression of the optimal Fo V angle. Then, the optimal FoV angle solution is used to solve the optimization PA. Since the PA problem has a non-convex form, we use an advanced technique of first-order Taylor approximation with difference of convex functions (D.C) method to solve it. Moreover, based on the Multi-Agent Deep Reinforcement Learning (MADRL) method, we develop a MADRL-based jamming mitigation algorithm to obtain the optimized solution of PA in near real-time. The numerical results show that the performance of the proposed MADRL-based jamming mitigation algorithm with low computational complexity is close to that of the optimization method.

We investigate the problem of dual protection for Warfighter Information Network-Tactical (WIN-T) of high-mobility ground combat vehicles (GCVs) against simultaneous energy-based and correlation-based interceptions. We design a joint resource optimization strategy in which the power allocation (PA) scheme controls transmit power, avoiding energy interception, and at the same time, the spreading factor assignment (SA) scheme manages correlation signal peaks to protect the network against the correlation analysis. We mathematically formulate this dual anti-interception resource allocation problem as a non-convex optimization model. We decompose this intractable optimization problem into two sub-problems, then solve the first sub-problem using an iterative method. To handle the non-convex form of the second sub-problem, we combine first-order Taylor approximation with the difference of convex functions (D.C) method. To obtain the optimized solution in near real-time, we propose a Multi-Agent Deep Reinforcement Learning (MADRL) approach. The numerical results show that the performance of the low computational complexity MADRL is close to that of the optimization method. Thus, the MADRL method has the potential to be applicable in high-complexity military scenarios.